[Nitro] Og.setup benchmarking results

Rob Pitt rob at motionpath.com
Wed Jan 11 04:32:52 EST 2006


It's worth keeping in mind you don't often start webapps so the startup
speed has little bearing on the actual CGI applications performance.

When Og does become a bottleneck in bigger web apps it's often not too
hard to convert just that bit to use more SQL.

On Wed, 2006-01-11 at 00:26 +0200, Aleksi Niemela wrote:
> Hello,
> 
> Tim was claiming Og.setup was too slow so I made a benchmark out of his 
> partial schema. 29 tables and some 70 constraints were created in 
> PostgreSQL. I measured total time, time spent at sql execution in ruby 
> postgres bindings (I use postgres-pr, the pure ruby lib) or in postgres 
> itself. Did the tests two times against a clean, freshly initiated 
> database and then two reruns.
> 
> Here are the results, hopefully they come through the mail.
> 
> 
> 	Fresh run 	
> 	
> 	
> 	Rerun 	
> 	
> 
> 	Total 	PSQL 	Og 	
> 	Total 	PSQL 	Og
> With sql printing 	18.16 	13.68 	4.48 	
> 	2.89 	0.49 	2.4
> 
> 	18.49 	14.04 	4.45 	
> 	2.86 	0.48 	2.38
> 
> 	
> 	
> 	
> 	
> 	
> 	
> 	
> Without sql printing 	
> 	
> 	
> 	
> 	
> 	
> 	
> 
> 	11.78 	9.83 	1.95 	
> 	2.26 	0.19 	2.07
> 
> 	11.56 	9.7 	1.86 	
> 	2.28 	0.19 	2.09
> 
> 
> I see no point trying to make Og side faster. Without debug prints Og 
> uses 20% of real time trying to generate SQL to execute in setup. In 
> rerun practically no SQL code executed so the same 2 secs gets spent in 
> Og. Maybe some portion of it could be shaved off. I didn't measure how 
> long will all the enchanting take out of that 2 secs.
> 
> Web app with complex Og schema won't have huge performance under CGI. 
> But I wouldn't expect so.
> 
>    - Aleksi
> 
> _______________________________________________
> Nitro-general mailing list
> Nitro-general at rubyforge.org
> http://rubyforge.org/mailman/listinfo/nitro-general




More information about the Nitro-general mailing list